It Got Wrecked.
In Part 1, I told you about Shannon’s Demon — the elegant idea that rebalancing between a volatile stock and cash can manufacture returns out of thin air, and the trio of geniuses (Claude Shannon, Ed Thorp, John Kelly) whose fingerprints are all over it. It’s one of those ideas that sounds so clean in theory that you start mentally spending the free money before you’ve tested it.
So I tested it. Twenty years of Nasdaq-100 (QQQ) price data, September 2006 through July 2026 — a window that happens to include the 2008 financial crisis, the 2020 COVID crash, and the 2022 rate-hike bear market, alongside one of the greatest tech bull runs in market history. If Shannon’s Demon was ever going to prove itself, this brutal-then-glorious stretch was the test.
Here’s the leaderboard, all measured over the same ~20-year period:
Look at that top row versus the Shannon’s Demon rows. The strategy named after one of the smartest men of the 20th century turned $10,000 into $47,000–$49,000, while simply buying and holding turned the same $10,000 into $181,000. Buying and holding beat the demon by roughly 4x.
Why the Demon Failed Here (And When It Actually Works)
This isn’t a case of “the math was wrong.” The math is fine — Shannon’s Demon genuinely does generate returns from rebalancing a sideways, choppy asset. The problem is that QQQ over this period wasn’t a sideways, choppy asset. It was a rocket ship with turbulence.
Every time the demon’s rebalancing rule kicked in, it was selling QQQ to buy cash on the way up (trimming a winner that kept winning) and buying QQQ with cash on the way down (adding to a position that, yes, eventually recovered — but the demon had already given away a chunk of the ride by holding 50% in cash the whole time). Constant rebalancing against a strongly trending asset isn’t harvesting volatility — it’s fighting the trend, over and over, for two decades straight.
The lesson here isn’t “Shannon’s Demon is fake.” It’s “the demon needs the right host.” It shines on assets with real volatility but no real drift. Feed it a compounding machine like the Nasdaq-100 instead, and it actively sabotages you.
Okay, But What About Smarter Exit Rules?
Fair question. Rebalancing is one way to try to manage risk — what about actual trend-following stop-losses? Sell when the trend breaks, sit in cash, buy back in when the trend resumes. Surely that beats doing nothing?
I tested a whole stable of these, all layered on top of a disciplined monthly $1,000 DCA into QQQ, same 20-year window:
Notice the pattern? Every single rule-based exit strategy underperformed just… buying every month and never selling. Not most of them. All of them. Some by a little, some catastrophically (that bottom-row Chandelier Exit strategy, with 144 stop-outs over 20 years, turned $239,000 into $390,000 — a strategy that just kept buying with zero rules turned the same money into $1.83 million).
There is a real trade-off buried in this table, though, and it’s worth naming honestly: every one of these rules did reduce max drawdown. The “do nothing” DCA strategy had to stomach a brutal -39.1% peak-to-trough decline at some point in the 20 years. The 10/20/50-week combo cut that all the way down to -16.7%. If you are the kind of investor who genuinely cannot stomach watching your account fall 39%, that’s not nothing — panic-selling at the real bottom is its own way to torch your returns. But it came at a steep price: less than half the final growth multiple.
The Whipsaw Problem
Why do these rules keep losing, even the “smart” ones? The short answer is whipsaw — the technical term for getting chopped up by false signals. A trend-following system sells when the trend breaks and buys back when it’s confirmed to have resumed. Sounds sensible. But “confirmed to have resumed” always happens after the price has already moved back up. Every full round-trip — sell, wait, confirm, buy back — quietly bleeds a little performance, and over 20 years and dozens of round trips, that bleed adds up to millions of dollars of difference.
There’s also a pattern hiding in the sell-count column: the more often a strategy pulls the trigger, the worse it tends to do. The twitchy 22-day Chandelier Exit fired 144 times and lost the most ground. The patient 200-day moving average fired only 25 times and lost the least. Trading less, it turns out, beat trading smarter.
This isn’t a fringe complaint about trend-following. Gary Antonacci, the investment professional who developed the widely-followed Dual Momentum strategy, has been candid that whipsaw — false signals that reverse course shortly after triggering — is simply a structural cost of any trend-following system, not a flaw specific to a badly-designed one. His own approach uses longer lookback periods specifically to reduce (not eliminate) that cost.
(Source: Better System Trader)
But there’s an even more specific reason these strategies keep losing, and it’s stranger — and more important — than simple whipsaw. It has to do with exactly which days the market goes up the most, and it explains why “time in the market” isn’t just a cliché on a coffee mug. That’s Part 3.
See exactly which days mattered most.
Next up: I found out exactly how many trading days out of roughly 5,000 actually mattered — and what happens to your returns if a trend-following strategy makes you miss even a handful of them. The number will make you rethink ever trying to time the market again.